Radeon X800PRO will beat NV40 Ultra

[991060]

Well, for me, Inq ain't that bad, people just need to think for a while before believing what Inq reported.

 

[Dave Baumann]

Either the the 3dMark03/ShaderMark figures I have seen are horribly wrong, or the 12-pipe pro won't be able to top the NV40 but will lag about 15% behind.

I'd estimate that they are wrong. There are not that many boards outside of ATI.

 

[Druga Runda]

But if the PRO is competitive to 6800U where does it leave the plain 6800 on 128bit bus? Aren't they the same price range?

128 bit bus???
6800 non ultra has a 256 bit bus.

Yes I think in my mind are still some early comments that it had 128 bit bus,

http://www.beyond3d.com/forum/viewtopic.php?t=11511


So well nevermind misinformed me as I haven't really caught up... but yes 256 bit bus makes much more sense.

 

[PaulS]

Counting down the minutes until the ATi fans come storming in with "I TOLD YOU SO !!!"

Whoops! Big surprise! A jack pops out. Then you laugh and the kids laugh and the dog laughs, and I DIE a little inside.

 

[jimmyjames123]

I'd estimate that they are wrong. There are not that many boards outside of ATI.

So you are implying that the 3dmark03/Shadermark scores for the X800 Pro actually are in line with the NV40 that you reviewed? Or not?

 

[PaulS]

I'd estimate that they are wrong. There are not that many boards outside of ATI.

So you are saying that the 3dmark03 scores for the X800 Pro actually are in line with the NV40 that you reviewed. How interesting

I like how a statement that suggests the figures are probably made up due to there not being many boards around, suddenly becomes confirmation that the Pro board will equal NV40U.

 

[jimmyjames123]

I'm not saying that this should be taken as fact, but DaveB's statement implies that the X800 Pro does not lag behind the NV40 (at least, the NV40 in it's A1 silicon form) with respect to 3dmark03/Shadermark. DaveB has been known to be cryptic in the past, so maybe I am not reading him right this time

Of course, there are plenty of other variables too. How high will the manufacturers clock their 6800U's? How much improvement will we see when the drivers mature? And so on...

 

[DemoCoder]

Either the the 3dMark03/ShaderMark figures I have seen are horribly wrong, or the 12-pipe pro won't be able to top the NV40 but will lag about 15% behind.

I'd estimate that they are wrong. There are not that many boards outside of ATI.

Ah, but are those that have them obeying their NDAs?

If X800 Pro is 500Mhz w/1200Mhz RAM as claimed by leak, that's a 25% clock advantage (but 33% less pipelines), with only 9% bandwidth advantage, so I am skeptical that it will outpace an NV40 by any significant margin in fillrate or shader throughput. Geometry wise, is another story.

Of course, the people msg'ing me have been plenty wrong before (about NV40 specs)

 

[DemoCoder]

BTW Dave, what did you mean when you said once "Register combiners are dead, long live register combiners!"

 

[jimmyjames123]

Right now, there is a lot of FUD being spread around, that is for sure! Should be a very interesting and exciting month

 

[tEd]

Either the the 3dMark03/ShaderMark figures I have seen are horribly wrong, or the 12-pipe pro won't be able to top the NV40 but will lag about 15% behind.

I'd estimate that they are wrong. There are not that many boards outside of ATI.

well i don't have one that's for sure

 

[991060]

BTW Dave, what did you mean when you said once "Register combiners are dead, long live register combiners!"

I see it as Dave was teasing NV3X's design.

 

[Maintank]

I think ATI coming out with the 800XT so soon after the X800 Pro indicates to me that we should not expect the Pro to beat the 6800 Ultra.

 

[jolle]

Either the the 3dMark03/ShaderMark figures I have seen are horribly wrong, or the 12-pipe pro won't be able to top the NV40 but will lag about 15% behind.

I'd estimate that they are wrong. There are not that many boards outside of ATI.

Ah, but are those that have them obeying their NDAs?

If X800 Pro is 500Mhz w/1200Mhz RAM as claimed by leak, that's a 25% clock advantage (but 33% less pipelines), with only 9% bandwidth advantage, so I am skeptical that it will outpace an NV40 by any significant margin in fillrate or shader throughput. Geometry wise, is another story.

Of course, the people msg'ing me have been plenty wrong before (about NV40 specs)

based on those rumours the fill rate is 6000 vs nv40 6400..
and IF the rumors that NV40 will use 475Mhz core is true that would be
7600 MP/Sec instead... and in the "special cases" such as doom3 Zpass
it would double that with 32pipes, or even (extremly speculative based on
anomaly in 1 benchmark at Xbits preview, very loose stuff) run 48 or 50x0, in possibly even more isolated cases perhaps...
remains to be seen if that "is anything" or not tho..

that wont matter at all unless other engines will use a similar approach
perhaps, so its one of those "maybe useful" features i guess..
some say its Stencil shadows in general that can benefit from it, dunno..
but disregarding it as a "safe speed advantage" until that is clear might
be a good idea.

Could possibly be compensated with higher shader performance in newer
games tho couldnt it? from ATIs side..
if they also use 2 units per pipelines, and.. um.. well "do it faster" what ever method they use to do it.. hehe..

older games run fast enough as it is, so the key would to put your card over
the top in the later titles, which in many cases would prolly end up with
shader peformance wouldnt it?
Even if one is slower in quake3 its still stellar FPS numbers..

NV40s open flank is the 8x FSAA performance, and a card that tends to get cpu limited in many situations has most of its current "value" tied up
in FSAA, AF and high resolution, so its a Point for ATI there if they show
good speed and quality in x8, if fillrate is notably lower then Nvidia they
might just loose that edge tho.. ehhh.. I would guess atleast..

Well, some more speculations and.. well questions..

 

[DoS]

Either the the 3dMark03/ShaderMark figures I have seen are horribly wrong, or the 12-pipe pro won't be able to top the NV40 but will lag about 15% behind.

I'd estimate that they are wrong. There are not that many boards outside of ATI.

Well, DC it seems you spilled some beans there...
Anyway, it doesn't matter if you are right or wrong we will know for sure in a few more days.
Is it just me or we knew many things more about NV40's expected performance a few days prior to the launch than we do about R420's now?? The info on R420 has been vague and many times contradicting (from "it will be enough slower for ATi to bring the XT part out after a month" to "it will be faster than 6800U"). One can only conclude that ATi is doing a hell of a job in the security department

 

[jolle]

they got some tight lid on their stuff at ATI..

or they just feed Inquirer alot of different storys so if there has been
a leak its been lost in all other bets on what kind of animal R420 is..

 

[jimmyjames123]

Both companies have done a reasonably good job this time around of keeping people guessing. I don't see that trend slowing anytime soon

 

[Moose]

I've already made the decision (last year) that I'm not upgrading my system until BTX PCI-E mainboards become available. In which case, I will buy a PCI-E card.

Have you seen any plans for PCI-E boards that are ATX form factor or are only BTX boards possible???


PCI-E information seems pretty scarce for such a big change happening so soon.

 

[digitalwanderer]

X800=15k 3dm2k3 default.

Any other questions?

 

[jimmyjames123]

I am assuming you are referring to the X800 XT, correct? Some other guy named lukar already mentioned the 15k 3dmark03 score.